Another mystery car
Discussion
borrani72 said:
It isn't 100% finished so there are a few details that need polishing, but the data could be used (in a more sophisticated CAD package) to produce a fully surfaced model for downstream processes.
Dr G said:
borrani72 said:
It isn't 100% finished so there are a few details that need polishing, but the data could be used (in a more sophisticated CAD package) to produce a fully surfaced model for downstream processes.
swisstoni said:
I would separate people’s interest in the mystery and the investigation from any great regard for the car itself.
I take your point. If the caption in the book had told us what the car was, this and other threads wouldn't exist. The mystery adds a great deal of cachet to the little car.However, even if it isn't something exotic, you have to admire the skill and dedication it must have required to build it. Assuming it's a home-built special, on an outdated and, at the time, rather undesirable base, I personaly think it's a very impressive piece of work.
borrani72 said:
My feeling is that it's probably an aluminium body.
Based purely upon the size of the wheel discs, the wheelbase is very close to the 78 inches (1981mm) of the Minor (plus or minus about half an inch). The track front and rear would then be 42 inches (1067mm).
Pretty small by modern standards, so finding the right (modern) running gear could be a challenge.
My main interest has always been the appearance, rather than the underpinnings. So something with a bespoke frame and later running gear would be one way to go. Better than cutting up a good pre-war Minor, unless there's a huge pile of spares available to work from.
Like many, I've been sucked down this rabbit hole.Based purely upon the size of the wheel discs, the wheelbase is very close to the 78 inches (1981mm) of the Minor (plus or minus about half an inch). The track front and rear would then be 42 inches (1067mm).
Pretty small by modern standards, so finding the right (modern) running gear could be a challenge.
My main interest has always been the appearance, rather than the underpinnings. So something with a bespoke frame and later running gear would be one way to go. Better than cutting up a good pre-war Minor, unless there's a huge pile of spares available to work from.
I've started again on the dimensional analysis, based upon the same image everyone else is using, albeit scaled-up with the highest available detail fidelity to enable some precise measurements.
The first finding is that there is some x-axis compression. The wheels in the image are not circular. The wheels in the image have a width:height ratio of 126:136 (front) and 123:136 (rear). The front wheel is notably turned to the driver's right, so is not a reliable scale for the image as a whole, so we'll use the rear for now.
Aligning the centre of the wheel trims (and assuming for these purposes that the slight turning of the front wheel doesn't destroy this calculation due to offsets / steering geometry), we can measure the wheelbase (1265 units on the same scale used above). If we accept the proposition from borrani72 that these wheel trims are 397mm in diameter, then the calculated wheelbase is 2041mm. Now if we factor in the likely small effect of the front wheel turning, it seems perfectly reasonable that the Sprite wheelbase of 2032 would fit. There is inherent imprecision here, but (assuming the wheel size is correct), I don't see this as short as 1981mm.
Are the tyres consistent with these dimensions? On the same scale, the tyre radius in the image is approximately 200/126*198.5 = 315mm, so a diameter of 630mm.
5.20x15 crossply tyres (common in period, now hard to get a hold of) have a median diameter of pretty much exactly 630mm. So the visible tyres are consistent with the other dimensions - which is all pitched off of the hub cap size.
If those were in fact 13 inch wheels, wearing (say) 5.20x13 crossplys (like a Sprite of the time would do), that scales the whole image by 584/630. That means a wheelbase of just 1890-1900mm (74" - 75" = 6'2" - 6'3"). That, to me, seems absurdly short.
Sticking with the 15 inch wheel hypothesis, we have a car with a wheelbase more or less the same as a Sprite's, with a scuttle area looking very much like a Sprite's, but wearing 15 inch wheels and matching tyres.
Now, look again at the front wheel arch - not only is the wheel high in the arch, it is very very long in the arch. There's almost no clearance, which is most unusual.
So, could this be a Sprite base running 15 inch wheels, then lowered to suit? Or it could be a bitsa achieved by grafting a Sprite-based nose onto an older chassis with a bespoke body?
Anyhow, that's enough from me for this moment. I'm working on some other theories right now, but not quite there yet
I know one member on this thread is adamant this car is Alpine based but Im not so sure, there are certainly elements of Alpine in the design but also lots of elements of other designs.
Although Ive been in the motor trade for 46 years one car Ive never worked on is an Alpine but to my knowledge they are monocoque not seperate chassis. Too my mind it would be much easier to build a bespoke body on a seperate chassis than radically redesign a monocoque.
Im also of the thought that it may be an aluminiun body as the work involved in making a pattern to take moulds from, then making a finished body tub and associated panels from fibreglass, would be an awful lot of work for a one off which this car seems to be.
Many small companies set out to build fibre glass kits for rebodying older vehicles and failed after the first prototype or after only manufacturing a couple of bodies, but these companies would have at least advertised at some point and someone somewhere would remember them or recognise there product but even all the experts on kit cars dont recognise this vehicle.
Although Ive been in the motor trade for 46 years one car Ive never worked on is an Alpine but to my knowledge they are monocoque not seperate chassis. Too my mind it would be much easier to build a bespoke body on a seperate chassis than radically redesign a monocoque.
Im also of the thought that it may be an aluminiun body as the work involved in making a pattern to take moulds from, then making a finished body tub and associated panels from fibreglass, would be an awful lot of work for a one off which this car seems to be.
Many small companies set out to build fibre glass kits for rebodying older vehicles and failed after the first prototype or after only manufacturing a couple of bodies, but these companies would have at least advertised at some point and someone somewhere would remember them or recognise there product but even all the experts on kit cars dont recognise this vehicle.
skwdenyer said:
Like many, I've been sucked down this rabbit hole.
I've started again on the dimensional analysis, based upon the same image everyone else is using, albeit scaled-up with the highest available detail fidelity to enable some precise measurements.
The first finding is that there is some x-axis compression. The wheels in the image are not circular. The wheels in the image have a width:height ratio of 126:136 (front) and 123:136 (rear). The front wheel is notably turned to the driver's right, so is not a reliable scale for the image as a whole, so we'll use the rear for now.
Aligning the centre of the wheel trims (and assuming for these purposes that the slight turning of the front wheel doesn't destroy this calculation due to offsets / steering geometry), we can measure the wheelbase (1265 units on the same scale used above). If we accept the proposition from borrani72 that these wheel trims are 397mm in diameter, then the calculated wheelbase is 2041mm. Now if we factor in the likely small effect of the front wheel turning, it seems perfectly reasonable that the Sprite wheelbase of 2032 would fit. There is inherent imprecision here, but (assuming the wheel size is correct), I don't see this as short as 1981mm.
Are the tyres consistent with these dimensions? On the same scale, the tyre radius in the image is approximately 200/126*198.5 = 315mm, so a diameter of 630mm.........................
....................Now, look again at the front wheel arch - not only is the wheel high in the arch, it is very very long in the arch. There's almost no clearance, which is most unusual.
An excellent post, thanks.I've started again on the dimensional analysis, based upon the same image everyone else is using, albeit scaled-up with the highest available detail fidelity to enable some precise measurements.
The first finding is that there is some x-axis compression. The wheels in the image are not circular. The wheels in the image have a width:height ratio of 126:136 (front) and 123:136 (rear). The front wheel is notably turned to the driver's right, so is not a reliable scale for the image as a whole, so we'll use the rear for now.
Aligning the centre of the wheel trims (and assuming for these purposes that the slight turning of the front wheel doesn't destroy this calculation due to offsets / steering geometry), we can measure the wheelbase (1265 units on the same scale used above). If we accept the proposition from borrani72 that these wheel trims are 397mm in diameter, then the calculated wheelbase is 2041mm. Now if we factor in the likely small effect of the front wheel turning, it seems perfectly reasonable that the Sprite wheelbase of 2032 would fit. There is inherent imprecision here, but (assuming the wheel size is correct), I don't see this as short as 1981mm.
Are the tyres consistent with these dimensions? On the same scale, the tyre radius in the image is approximately 200/126*198.5 = 315mm, so a diameter of 630mm.........................
....................Now, look again at the front wheel arch - not only is the wheel high in the arch, it is very very long in the arch. There's almost no clearance, which is most unusual.
You certainly gave my little grey cells a workout trying to understand why we are getting different results from essentially the same calculation.
Firstly, I re-checked my own calculations.
To get the maximum resolution, I opened the image in a graphics package and drew some rectangles.
This allowed me to enlarge the wheels to fill the screen and measured the diameters (horizontally), then to scroll across to measure the wheelbase in sections, from each rectangle to the next (to be added together later) to give the most accurate result possible. At this magnification, the limiting factor are the coloured dots of the printing process, so we can get no closer.
Here are the measurements I got.....
With this, I did the four calculations shown, using the leading-edge measurement of the wheelbase, then the trailing-edge measurement (see top diagramme, above), each combined with the two slightly different wheel disc sizes from the image.
The results came out between 1981mm and 1992mm for the full-size wheelbase (the pre-war Morris Minor is 1981mm).
Then I had to try to explain why we are getting different results.
I think the first clue is what you said about the width/height ratio of the wheels. I noticed that, as we can't see the lower edge of the rim, you must be using the radius, not the diameter.
As we are looking down slightly on the car, the centre of the wheel disc (being raised, and therefore closer to us) will be subject to parallax. The measurement on the photograph from the hub-centre to the top of the wheel rim will be greater than that to the bottom of the rim, as measured on the image. The pink lines run the same length from the centre, but finish at different points on the disc.
I think this is the explanation for any seeming distortion on the x-axis that you are finding [although I also think there is radial distortion in the image (apparantly common in 1960s zoom lenses). This seems to work in our favour here, as it seems to have equalised the width (diameter) of the wheel discs in the image, making them almost the same size. (Radial distortion tends to reduce the size of objects closer to the centre of the image, so the rear wheel is slightly enlarged, being toward the outside edge of the picture, and therefore largely cancelling out any perspective regression). This is also why I used special software to correct the distortion before I created the CAD model].
Where parallax matters in the calculation of the wheelbase is at the rear wheel.
If you measure the centre of the wheel to the forward and then to the trailing edge of the rim, the results are different. (Again, the lines in the image are the same length).
To allow for this difference, I measured and deducted 6.5mm from the screen measurement and repeated one of the calculations... 397/(300.5-6.5) x 1507 equals 2035 . The result was this full-size wheelbase of 2035mm – very close to your 2041mm result.
If you measure across the full width of the rims, then everything will be on the same plane, and remove any parallax errors. You can measure the wheelbase from trailing edge to trailing edge (or leading edges), rather than wheel centres. If I'm correct, I suspect that we will then get near identical results.
…..........................................................................................................................................................
The outer diameter of the tyres in my model is 634mm.
The apparent lack of clearance around the front wheels is accounted for by the overhang of the body, which is somewhat wider than the tyres. This allows for a normal amount of steering lock.
Edited by borrani72 on Friday 21st May 20:01
Edited by borrani72 on Friday 21st May 20:19
borrani72 said:
An excellent post, thanks.
You certainly gave my little grey cells a workout trying to understand why we are getting different results from essentially the same calculation.
LOLYou certainly gave my little grey cells a workout trying to understand why we are getting different results from essentially the same calculation.
borrani72 said:
Snipped a bunch of interesting stuff...
As we are looking down slightly on the car, the centre of the wheel disc (being raised, and therefore closer to us) will be subject to parallax. The measurement on the photograph from the hub-centre to the top of the wheel rim will be greater than that to the bottom of the rim, as measured on the image. The pink lines run the same length from the centre, but finish at different points on the disc.
I think this is the explanation for any seeming distortion on the x-axis that you are finding [although I also think there is radial distortion in the image (apparantly common in 1960s zoom lenses). This seems to work in our favour here, as it seems to have equalised the width (diameter) of the wheel discs in the image, making them almost the same size. (Radial distortion tends to reduce the size of objects closer to the centre of the image, so the rear wheel is slightly enlarged, being toward the outside edge of the picture, and therefore largely cancelling out any perspective regression). This is also why I used special software to correct the distortion before I created the CAD model].
Good points. The distortion isn't just about the original lens. This image seems to have been cropped off-centre. It was probably cropped as a print, before being re-photographed using a rostrum camera. So we're going to be looking at a compound effect of multiple lenses.As we are looking down slightly on the car, the centre of the wheel disc (being raised, and therefore closer to us) will be subject to parallax. The measurement on the photograph from the hub-centre to the top of the wheel rim will be greater than that to the bottom of the rim, as measured on the image. The pink lines run the same length from the centre, but finish at different points on the disc.
I think this is the explanation for any seeming distortion on the x-axis that you are finding [although I also think there is radial distortion in the image (apparantly common in 1960s zoom lenses). This seems to work in our favour here, as it seems to have equalised the width (diameter) of the wheel discs in the image, making them almost the same size. (Radial distortion tends to reduce the size of objects closer to the centre of the image, so the rear wheel is slightly enlarged, being toward the outside edge of the picture, and therefore largely cancelling out any perspective regression). This is also why I used special software to correct the distortion before I created the CAD model].
And we've got parallax
borrani72 said:
Where parallax matters in the calculation of the wheelbase is at the rear wheel.
If you measure the centre of the wheel to the forward and then to the trailing edge of the rim, the results are different. (Again, the lines in the image are the same length).
OK, so I've fired up Photoshop & used the lens correction capability to correct the image. As it stands, the image is clearly a crop - the barrel distortion is not centred on the centre of the image. I've attempted to compensate by shifting the image on a larger canvas, then corrected for barrel distortion and vertical perspective.If you measure the centre of the wheel to the forward and then to the trailing edge of the rim, the results are different. (Again, the lines in the image are the same length).
Now my rear wheel is basically circular
The problem? The centre isn't in the centre. Why's that? Because those hub caps have a lot of depth to them. We're looking down and off-angle at them. So parallax means that the visible centre of the hub cap isn't actually in the centre of the wheel. As your pink lines are showing you. As an example for others, consider this image:
The centre of the hub cap isn't in the centre of the hub cap (IYSWIM). Even if you stretch the image to make the hub cap circular, you won't get the centre of the centre (as it were) to be in the centre of the circle.
It is that problem we have here. The lighting of the rear wheel doesn't help, either.
Correcting for that (i.e. using the geometric centre of the hub cap, not the centre of the centre), using the horizontal diameter as a reference, and using trailing edge to trailing edge as the measure of the wheelbase, for the same 397mm hub cap I get a wheelbase of 1379 / 282 * 379 = 1853mm.
Yes, that's a huge difference from my original measurement. So I perform the same analysis in the same way on the original image from my first post (i.e. without correction for the barrel distortion & perspective). What do I get? 2037mm.
Why the huge difference? Because the 2 images I'm using have been handled in different ways. The latest image I used was the highest-resolution scan I could find of the original, progressively enlarged by me to preserve detail. I'm pretty confident in the edge identification. The original image I used was from a different thread post, and is clearly of poorer quality. I've also boosted contrast & exposure compensation to try to find the edges, but at this resolution the problem is that leads to some "spreading" of highlights. Since the whole thing is based off of the diameter of a wheel disc, those tiny differences have a disproportionate effect on the overall calculation.
So, what to do? Get smarter
So I've used Super Resolution Imaging, an AI-enhanced technique for enlarging images to do some of what CSI etc used to pretend was possible 20 years ago & enlarge the images whilst restoring some of the lost detail. The idea is that if the underlying image is bigger (more pixels), the size of the error is smaller.What's the result? Calculated in the same way (geometric centres of the wheel discs, measured centre-to-centre), I get 2007mm.
So now I'm clear as mud on where we are. I've got a copy of the book on order (I know, the rabbit hole becomes a warren now), I'm going to get the best possible scan I can from it & see if I can come to any better conclusions. But from what I see, the print quality of the original is sufficiently poor that precise measurements may ultimately be impossible due to the limitations of the printing process.
The only thing I'm reasonably confident of is that that the wheelbase is somewhere between 1850 and 2050 mm, but a tolerance of 11% isn't a terribly great achievement!
And all of that presupposes that we really do have 397mm diameter wheel covers. If those covers are actually 387mm (say), the wheelbase calculation will be out still further...All of this is a good example for the reader as to why even apparently simple stuff is pretty hard. Even something as apparently simple as working out how big a circle on a picture is really isn't that simple at all.
Out of simple interest, using my original calculation as a guide, I calculated the potential length of the vehicle. I get about 3.5m, which is basically identical to an MG Midget.
Anyhow, I'll wait for my book to arrive before doing more on this, as I don't even know how good the original scans we're working from are!
jules_s said:
For some time I've thought it is/was Austin Healey based in some way
There seem to be quite a few Sebring/WSM cues in the design - that Naca duct behind the window...
The problem is there are so many different cues on this car, that you can see echoes in all sorts of vehicles from the period.There seem to be quite a few Sebring/WSM cues in the design - that Naca duct behind the window...
In terms of overall form & hooded headlights, the Allemano DB2-4 from 1953 is a dead ringer (scale it in Photoshop and it is almost a perfect match to the silhouette of this car) - but did Allemano (or the credited designer, Sergio Sartorelli) do something smaller? Or was it just inspiration?
The Ghia-Aigle Lotus Eleven is also a good match - the door treatment, the wraparound screen, the proportions. That was a Michelotti design. Did he do a Sprite, say? Without going to the Michelotti archive, I've no idea - he did *so many* one-offs for individual customers, that keep emerging out of the woodwork.
The surfacing of the mystery car looks to me like a mixture of GFRP and aluminium. Who did that? Well, if you look at, say, Peter Ecury's "mystery Sprite" you'll see original steel, GFRP and aluminium muddled together to achieve the overall look (see below).
Nobody knows who made that one right now, either (the article above gives the plate incorrectly - the real 347RAL was actually built by Lenhams for Tony Pay, and is not apparently the same car - but they look very similar!)
Many like to think the mystery car a product of Williams & Pritchard. That may be so, and if it is so we may never really know; as Mr. Pritchard, in an interview in 1989, said: "All our jigs went years ago. And we did not keep any photographic records, either. We were so busy and involved, it just did not occure to us. So I am having to go by memory. But things happened so fast, that my memories are jumbled. It never occured to us, that what we were making, would one day become classics."
The best hope we have is of getting just *one* step further down the line, and hoping that will give us another avenue to explore.
skwdenyer said:
The distortion isn't just about the original lens. This image seems to have been cropped off-centre. It was probably cropped as a print, before being re-photographed using a rostrum camera. So we're going to be looking at a compound effect of multiple lenses.
And we've got parallax
OK, so I've fired up Photoshop & used the lens correction capability to correct the image. As it stands, the image is clearly a crop - the barrel distortion is not centred on the centre of the image. I've attempted to compensate by shifting the image on a larger canvas, then corrected for barrel distortion and vertical perspective.
................for the same 397mm hub cap I get a wheelbase of 1379 / 282 * 379 = 1853mm.
Which image are you starting with? A lot of images on the thread have been cropped again.And we've got parallax
OK, so I've fired up Photoshop & used the lens correction capability to correct the image. As it stands, the image is clearly a crop - the barrel distortion is not centred on the centre of the image. I've attempted to compensate by shifting the image on a larger canvas, then corrected for barrel distortion and vertical perspective.
................for the same 397mm hub cap I get a wheelbase of 1379 / 282 * 379 = 1853mm.
Below is the full image as it appears in the 1967 book (the 2008 reprint is cropped very slightly more at the bottom edge).
(Which edition has the best image to scan from, I don't know. If the reprint was scanned from an original book (quite likely) then it won't be as good, though it may have come from the original artwork (less likely, perhaps)?).
The horizon is virtually at the centre of the image. Assuming the height is not cropped, or is minimally cropped, then the standard 3:2 film ratio would be wider - to the dashed pencil line at the top left. Whether cropped left, right or both is difficult to say, but the verticals at the far right of the picture seem to have slightly less vertical perspective than those on the left, which could suggest the right side is cropped.
In your calculation above, exactly where are the numbers (1379 / 282 * 379 = 1853mm) from? Should it not be 397, not 379 (the wheel disc diameter)?
The simplest approach to the distortion seems to me to calculate the wheelbase from the front wheel, and then from the rear. The distortion has almost removed the perspective regression, and the two figures are therefore very close to one another. The true figure, logically, must be between the two (unless the radial distortion is massive, which I think would be noticeable). This, I think, gives a good sense of the likely tolerance of the wheelbase measurement, as errors will tend to cancel-out.
For this reason, I only corrected the radial distortion for the CAD model, not for the wheelbase calculation. This (reduction of radial distortion) was done by eye, matching the perspective regression on the wheel sizes of the CAD model against the photo'.
In fact, to get a good idea of the wheelbase, I would suggest that INTRODUCING radial distortion to make the wheels the same size would be a good way to reduce the perspective regression, had it not happened here by chance.
Edited by borrani72 on Saturday 22 May 02:22
borrani72 said:
Hi Tony,
the best scan I have found seems to be around 400 dpi, so that sounds like it could be very useful. Thanks.
Would it keep the resolution when posted here?
I'll upload the scan to my website and provide a download link instead of uploading the image here.the best scan I have found seems to be around 400 dpi, so that sounds like it could be very useful. Thanks.
Would it keep the resolution when posted here?
I'll also scan it in .bmp format so there are no .jpg compression artefacts.
Watch this space as they say!
Gassing Station | Classic Cars and Yesterday's Heroes | Top of Page | What's New | My Stuff